pith. sign in

arxiv: 2605.14712 · v1 · pith:T3KETLZKnew · submitted 2026-05-14 · 💻 cs.RO · cs.AI· cs.CL· cs.CV

IntentVLA: Short-Horizon Intent Modeling for Aliased Robot Manipulation

Shijie Lian , Bin Yu , Xiaopeng Lin , Zhaolong Shen , Laurence Tianruo Yang , Yurun Jin , Haishan Liu , Changti Wu
show 3 more authors
Hang Yuan Cong Huang Kai Chen
This is my paper

Pith reviewed 2026-06-30 20:53 UTC · model grok-4.3

classification 💻 cs.RO cs.AIcs.CLcs.CV
keywords robot manipulationvisual-language-actionintent modelingobservation aliasingimitation learningaction chunkinghistory conditioning
0
0 comments X

The pith

Encoding recent visual observations into a short-horizon intent representation allows visual-language-action policies to generate consistent action chunks under observation aliasing.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Robot imitation datasets often contain multiple valid action sequences for the same visual-language input because demonstrators pursue different short-term goals. Standard policies that condition only on the current frame can switch intents between replanning steps, producing jerky or failed executions. IntentVLA extracts a compact summary of recent history to represent the current short-horizon intent and uses that summary to guide each new action chunk. The approach is tested on a new benchmark that isolates short-horizon aliasing as well as on established robot manipulation suites. If the method works, policies can maintain intent consistency across replans without extra sensors or explicit state estimation.

Core claim

IntentVLA is a history-conditioned VLA framework that encodes recent visual observations into a compact short-horizon intent representation and conditions chunk generation on this representation, which improves rollout stability and outperforms baselines on AliasBench, SimplerEnv, LIBERO, and RoboCasa.

What carries the argument

The short-horizon intent representation, a compact encoding of recent visual observations that disambiguates the current task phase or intent for conditioning action generation.

If this is right

  • Policies achieve more stable rollouts by avoiding inter-chunk intent conflicts.
  • Performance gains hold across multiple simulation environments and benchmarks designed for aliasing.
  • The framework can be applied to existing VLA architectures by adding the history encoder and conditioning.
  • Training remains feasible without introducing new inconsistencies.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar history-based intent modeling could help in partially observable real-world settings where visual aliasing is common.
  • The method might reduce the need for high-frequency replanning if intent consistency is maintained.
  • Extending the intent representation to include language or proprioceptive history could further improve disambiguation.

Load-bearing premise

The multimodal nature of imitation data stems mainly from different short-horizon intents that recent observations can summarize compactly enough for conditioning to resolve conflicts.

What would settle it

A controlled experiment where the intent-conditioned model is compared to the baseline on tasks with known intent switches, measuring if the frequency of action chunk conflicts decreases significantly.

Figures

Figures reproduced from arXiv: 2605.14712 by Bin Yu, Changti Wu, Cong Huang, Haishan Liu, Hang Yuan, Kai Chen, Laurence Tianruo Yang, Shijie Lian, Xiaopeng Lin, Yurun Jin, Zhaolong Shen.

Figure 1
Figure 1. Figure 1: An illustrative example of short-horizon intent ambiguity under frame-only conditioning. The task is ordinary: the robot puts a piece of bread into a skillet for cooking and then returns it to the plate. The ambiguity appears because similar bread-in-gripper observations occur before two different continuations: placing the bread into the skillet and returning it to the plate. A frame-conditioned chunk pol… view at source ↗
Figure 2
Figure 2. Figure 2: Representative observation aliasing patterns in AliasBench. The quantitative observation-aliasing diagnostic is shown in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Quantitative observation-aliasing diagnostic on AliasBench. Back-and-Forth uses intra-episode re￾trieval with a 20-frame temporal gap; all other families use cross-episode retrieval. The diagnostic is not a policy success metric. Instead, it measures whether visually nearby states in the ambiguity window can correspond to different next intents. Left: roughly half of the top-k neighbors (k = 5) come from a… view at source ↗
Figure 4
Figure 4. Figure 4: Overview of IntentVLA. A Qwen3-VL backbone encodes the current image and language instruction, while a frozen VGGT-1B history encoder extracts recent visual evidence. IntentVLA fuses the history tokens with the current visual-language context through gated cross-attention, appends a compact short-horizon intent token, and conditions a DiT-based flow-matching action head for chunk generation. and predicts a… view at source ↗
Figure 5
Figure 5. Figure 5: Inter-chunk consistency in AliasBench ambiguity windows. We compare IntentVLA against the strongest feasible history-as-context baseline in [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
read the original abstract

Robot imitation data are often multimodal: similar visual-language observations may be followed by different action chunks because human demonstrators act with different short-horizon intents, task phases, or recent context. Existing frame-conditioned VLA policies infer each chunk from the current observation and instruction alone, so under partial observability they may resample different intents across adjacent replanning steps, leading to inter-chunk conflict and unstable execution. We introduce IntentVLA, a history-conditioned VLA framework that encodes recent visual observations into a compact short-horizon intent representation and uses it to condition chunk generation. We further introduce AliasBench, a 12-task ambiguity-aware benchmark on RoboTwin2 with matched training data and evaluation environments that isolate short-horizon observation aliasing. Across AliasBench, SimplerEnv, LIBERO, and RoboCasa, IntentVLA improves rollout stability and outperforms strong VLA baselines

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The paper introduces IntentVLA, a history-conditioned VLA policy that encodes recent visual observations into a compact short-horizon intent representation used to condition action chunk generation. The goal is to reduce inter-chunk conflicts caused by multimodal aliasing in imitation data under partial observability. The authors also present AliasBench, a 12-task benchmark on RoboTwin2 designed to isolate short-horizon observation aliasing with matched training and evaluation data. Empirical results are reported showing improved rollout stability and outperformance versus strong VLA baselines across AliasBench, SimplerEnv, LIBERO, and RoboCasa.

Significance. If the reported gains hold under the full experimental protocol, the work supplies a practical mechanism for stabilizing chunk-based VLA execution without requiring full history or additional sensors. AliasBench provides a controlled testbed for aliasing phenomena that are otherwise difficult to isolate, which could support follow-on research. The approach is incremental on existing VLA architectures yet directly targets a load-bearing source of execution instability in real-robot deployment.

minor comments (2)
  1. [Abstract] The abstract states performance improvements but supplies no numerical values, metrics, or effect sizes; adding one or two headline numbers (e.g., success-rate deltas on AliasBench) would improve immediate readability.
  2. [Methods] The description of the intent encoder architecture and its training objective is referenced but not expanded in the provided excerpt; ensure the methods section supplies the precise input window length, embedding dimension, and loss formulation so that the compactness claim can be verified.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive review, the recognition of IntentVLA's practical contribution to stabilizing chunk-based VLA policies under aliasing, and the recommendation for minor revision. We appreciate the note on AliasBench as a controlled testbed and will incorporate any minor suggestions in the revised manuscript.

Circularity Check

0 steps flagged

No significant circularity; empirical method without derivation chain

full rationale

The provided manuscript text consists of an empirical proposal for a history-conditioned VLA policy and an associated benchmark (AliasBench). No equations, derivations, fitted parameters presented as predictions, or load-bearing self-citations appear in the abstract or described structure. The central claim is that conditioning on a short-horizon intent encoder improves stability; this is evaluated via rollout metrics on external suites rather than reducing to a self-definition or renamed input. The paper is self-contained against its benchmarks with no visible reduction of outputs to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no free parameters, axioms, or invented entities are described.

pith-pipeline@v0.9.1-grok · 5718 in / 1138 out tokens · 34800 ms · 2026-06-30T20:53:53.195980+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

55 extracted references · 31 canonical work pages · 17 internal anchors

  1. [1]

    Hongzhe Bi, Lingxuan Wu, Tianwei Lin, Hengkai Tan, Zhizhong Su, Hang Su, and Jun Zhu. 2025. H-RDT: Human manipulation enhanced bimanual robotic manipulation.arXiv preprint arXiv:2507.23523

    work page arXiv 2025

  2. [2]

    Johan Bjorck, Fernando Castañeda, Nikita Cherniadev, Xingye Da, Runyu Ding, Linxi Fan, Yu Fang, Dieter Fox, Fengyuan Hu, Spencer Huang, and 1 others. 2025. GR00T N1: An open foundation model for generalist humanoid robots.arXiv preprint arXiv:2503.14734

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  3. [3]

    $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control

    Kevin Black, Noah Brown, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, Lachy Groom, Karol Hausman, Brian Ichter, Szymon Jakubczak, Tim Jones, Liyiming Ke, Sergey Levine, Adrian Li-Bell, Mohith Mothukuri, Suraj Nair, Karl Pertsch, Lucy Xiaoyang Shi, and 5 others. 2024.π 0: A vision- language-action flow model for general robot cont...

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  4. [4]

    Qingwen Bu, Jisong Cai, Li Chen, Xiuqi Cui, Yan Ding, Siyuan Feng, Shenyuan Gao, Xindong He, Xu Huang, Shu Jiang, and 1 others. 2025. Agibot world colosseo: A large-scale manipulation platform for scalable and intelligent embodied systems.arXiv preprint arXiv:2503.06669

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    Qingwen Bu, Yanting Yang, Jisong Cai, Shenyuan Gao, Guanghui Ren, Maoqing Yao, Ping Luo, and Hongyang Li. 2025. Univla: Learning to act anywhere with task-centric latent actions.arXiv preprint arXiv:2505.06111

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  6. [6]

    Tianxing Chen, Zanxin Chen, Baijun Chen, Zijian Cai, Yibin Liu, Zixuan Li, Qiwei Liang, Xianliang Lin, Yiheng Ge, Zhenyu Gu, and 1 others. 2025. Robotwin 2.0: A scalable data generator and benchmark with strong domain randomization for robust bimanual robotic manipulation.arXiv preprint arXiv:2506.18088

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  7. [7]

    Tianxing Chen, Yuran Wang, Mingleyang Li, Yan Qin, Hao Shi, Zixuan Li, Yifan Hu, Yingsheng Zhang, Kaixuan Wang, Yue Chen, and 1 others. 2026. Rmbench: Memory-dependent robotic manipulation benchmark with insights into policy design.arXiv preprint arXiv:2603.01229

    work page arXiv 2026

  8. [8]

    Yi Chen, Yuying Ge, Hui Zhou, Mingyu Ding, Yixiao Ge, and Xihui Liu. 2026. Dial: Decoupling intent and action via latent world modeling for end-to-end vla.arXiv preprint arXiv:2603.29844

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  9. [9]

    StarVLA Community. 2026. Starvla: A lego-like codebase for vision-language-action model developing.arXiv preprint arXiv:2604.05014

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  10. [10]

    GEAR-Team, Allison Azzolini, Johan Bjorck, Valts Blukis, Fernando Castañeda, Rahul Chand, and 1 others

  11. [11]

    nvidia.com/labs/gear/gr00t-n1_6/

    Gr00t n1.6: An improved open foundation model for generalist humanoid robots.https://research. nvidia.com/labs/gear/gr00t-n1_6/

  12. [12]

    Renming Huang, Chendong Zeng, Wenjing Tang, Jintian Cai, Cewu Lu, and Panpan Cai. 2026. Mimic intent, not just trajectories.arXiv preprint arXiv:2602.08602

    work page arXiv 2026

  13. [13]

    $\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization

    Physical Intelligence, Kevin Black, Noah Brown, James Darpinian, Karan Dhabalia, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, Manuel Y . Galliker, Dibya Ghosh, Lachy Groom, Karol Hausman, Brian Ichter, Szymon Jakubczak, Tim Jones, Liyiming Ke, Devin LeBlanc, and 17 others. 2025.π0.5: a vision-language-action model with open-world...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  14. [14]

    Alexander Khazatsky, Karl Pertsch, Suraj Nair, Ashwin Balakrishna, Sudeep Dasari, Siddharth Karamcheti, Soroush Nasiriany, Mohan Kumar Srirama, Lawrence Yunliang Chen, Kirsty Ellis, and 1 others. 2024. Droid: A large-scale in-the-wild robot manipulation dataset.arXiv preprint arXiv:2403.12945

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  15. [15]

    Moo Jin Kim, Chelsea Finn, and Percy Liang. 2025. Fine-tuning vision-language-action models: Optimizing speed and success.arXiv preprint arXiv:2502.19645

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  16. [16]

    Moo Jin Kim, Karl Pertsch, Siddharth Karamcheti, Ted Xiao, Ashwin Balakrishna, Suraj Nair, Rafael Rafailov, Ethan P Foster, Pannag R Sanketi, Quan Vuong, Thomas Kollar, Benjamin Burchfiel, Russ Tedrake, Dorsa Sadigh, Sergey Levine, Percy Liang, and Chelsea Finn. 2024. OpenVLA: An open-source vision-language- action model. InConference on Robot Learning (CoRL)

    2024

  17. [17]

    Chengmeng Li, Junjie Wen, Yaxin Peng, Yan Peng, and Yichen Zhu. 2026. Pointvla: Injecting the 3d world into vision-language-action models.IEEE Robotics and Automation Letters, 11(3):2506–2513

    2026

  18. [18]

    Fuhao Li, Wenxuan Song, Han Zhao, Jingbo Wang, Pengxiang Ding, Donglin Wang, Long Zeng, and Haoang Li. 2025. Spatial forcing: Implicit spatial representation alignment for vision-language-action model.arXiv preprint arXiv:2510.12276. 13

    work page arXiv 2025

  19. [19]

    Peiyan Li, Yixiang Chen, Hongtao Wu, Xiao Ma, Xiangnan Wu, Yan Huang, Liang Wang, Tao Kong, and Tie- niu Tan. 2025. Bridgevla: Input-output alignment for efficient 3d manipulation learning with vision-language models. InAdvances in neural information processing systems (NeurIPS)

    2025

  20. [20]

    Qixiu Li, Yaobo Liang, Zeyu Wang, Lin Luo, Xi Chen, Mozheng Liao, Fangyun Wei, Yu Deng, Sicheng Xu, Yizhong Zhang, and 1 others. 2024. CogACT: A foundational vision-language-action model for synergizing cognition and action in robotic manipulation.arXiv preprint arXiv:2411.19650

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  21. [21]

    Xinghang Li, Peiyan Li, Minghuan Liu, Dong Wang, Jirong Liu, Bingyi Kang, Xiao Ma, Tao Kong, Hanbo Zhang, and Huaping Liu. 2024. Towards generalist robot policies: What matters in building vision-language- action models.arXiv preprint arXiv:2412.14058

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  22. [22]

    Xuanlin Li, Kyle Hsu, Jiayuan Gu, Oier Mees, Karl Pertsch, Homer Rich Walke, Chuyuan Fu, Ishikaa Lu- nawat, Isabel Sieh, Sean Kirmani, Sergey Levine, Jiajun Wu, Chelsea Finn, Hao Su, Quan Vuong, and Ted Xiao. 2024. SimplerEnv: Evaluating real-world robot manipulation policies in simulation. InConference on Robot Learning (CoRL)

    2024

  23. [23]

    Shijie Lian, Bin Yu, Xiaopeng Lin, Laurence T Yang, Zhaolong Shen, Changti Wu, Yuzhuo Miao, Cong Huang, and Kai Chen. 2026. Langforce: Bayesian decomposition of vision language action models via latent action queries.arXiv e-prints, pages arXiv–2601

    2026

  24. [24]

    Tao Lin, Gen Li, Yilei Zhong, Yanwen Zou, Yuxin Du, Jiting Liu, Encheng Gu, and Bo Zhao. 2025. Evo-0: Vision-language-action model with implicit spatial understanding.arXiv preprint arXiv:2507.00416

    work page arXiv 2025

  25. [25]

    Xiaopeng Lin, Shijie Lian, Bin Yu, Ruoqi Yang, Changti Wu, Yuzhuo Miao, Yurun Jin, Yukun Shi, Cong Huang, Bojun Cheng, and 1 others. 2025. PhysBrain: Human egocentric data as a bridge from vision language models to physical intelligence.arXiv preprint arXiv:2512.16793

    work page arXiv 2025

  26. [26]

    Bo Liu, Yifeng Zhu, Chongkai Gao, Yihao Feng, Qiang Liu, Yuke Zhu, and Peter Stone. 2023. LIBERO: Benchmarking knowledge transfer for lifelong robot learning.Advances in neural information processing sys- tems (NeurIPS), 36:44776–44791

    2023

  27. [27]

    Songming Liu, Bangguo Li, Kai Ma, Lingxuan Wu, Hengkai Tan, Xiao Ouyang, Hang Su, and Jun Zhu

  28. [28]

    Rdt2: Exploring the scaling limit of umi data towards zero-shot cross-embodiment generalization.arXiv preprint arXiv:2602.03310

    work page arXiv

  29. [29]

    Songming Liu, Lingxuan Wu, Bangguo Li, Hengkai Tan, Huayu Chen, Zhengyi Wang, Ke Xu, Hang Su, and Jun Zhu. 2025. RDT-1b: a diffusion foundation model for bimanual manipulation. InInternational Conference on Learning Representations (ICLR)

    2025

  30. [30]

    Ilya Loshchilov and Frank Hutter. 2017. Decoupled weight decay regularization. InInternational Conference on Learning Representations (ICLR)

    2017

  31. [31]

    Yulin Luo, Hao Chen, Zhuangzhe Wu, Bowen Sui, Jiaming Liu, Chenyang Gu, Zhuoyang Liu, Qiuxuan Feng, Jiale Yu, Shuo Gu, and 1 others. 2026. Look before acting: Enhancing vision foundation representations for vision-language-action models.arXiv preprint arXiv:2603.15618

    work page arXiv 2026

  32. [32]

    Soroush Nasiriany, Abhiram Maddukuri, Lance Zhang, Adeet Parikh, Aaron Lo, Abhishek Joshi, Ajay Man- dlekar, and Yuke Zhu. 2024. RoboCasa: Large-scale simulation of everyday tasks for generalist robots. In Robotics: Science and Systems

    2024

  33. [33]

    Abby O’Neill, Abdul Rehman, Abhiram Maddukuri, Abhishek Gupta, Abhishek Padalkar, Abraham Lee, Acorn Pooley, Agrim Gupta, Ajay Mandlekar, Ajinkya Jain, and 1 others. 2024. Open x-embodiment: Robotic learning datasets and rt-x models: Open x-embodiment collaboration. In2024 IEEE International Conference on Robotics and Automation (ICRA), pages 6892–6903. IEEE

    2024

  34. [34]

    Karl Pertsch, Kyle Stachowicz, Brian Ichter, Danny Driess, Suraj Nair, Quan Vuong, Oier Mees, Chelsea Finn, and Sergey Levine. 2025. FAST: Efficient action tokenization for vision-language-action models.arXiv preprint arXiv:2501.09747

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  35. [35]

    Delin Qu, Haoming Song, Qizhi Chen, Yuanqi Yao, Xinyi Ye, Yan Ding, Zhigang Wang, JiaYuan Gu, Bin Zhao, Dong Wang, and 1 others. 2025. Spatialvla: Exploring spatial representations for visual-language-action model.arXiv preprint arXiv:2501.15830

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  36. [36]

    Jeff Rasley, Samyam Rajbhandari, Olatunji Ruwase, and Yuxiong He. 2020. Deepspeed: System optimiza- tions enable training deep learning models with over 100 billion parameters. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining, pages 3505–3506. 14

    2020

  37. [37]

    Yichao Shen, Fangyun Wei, Zhiying Du, Yaobo Liang, Yan Lu, Jiaolong Yang, Nanning Zheng, and Baining Guo. 2025. VideoVLA: Video generators can be generalizable robot manipulators. InAdvances in neural information processing systems (NeurIPS)

    2025

  38. [38]

    Hao Shi, Bin Xie, Yingfei Liu, Lin Sun, Fengrong Liu, Tiancai Wang, Erjin Zhou, Haoqiang Fan, Xiangyu Zhang, and Gao Huang. 2026. Memoryvla: Perceptual-cognitive memory in vision-language-action models for robotic manipulation. InInternational Conference on Learning Representations (ICLR)

    2026

  39. [39]

    Jingwen Sun, Wenyao Zhang, Zekun Qi, Shaojie Ren, Zezhi Liu, Hanxin Zhu, Guangzhong Sun, Xin Jin, and Zhibo Chen. 2026. Vla-jepa: Enhancing vision-language-action model with latent world model.arXiv preprint arXiv:2602.10098

    work page arXiv 2026

  40. [40]

    Octo Model Team, Dibya Ghosh, Homer Walke, Karl Pertsch, Kevin Black, Oier Mees, Sudeep Dasari, Joey Hejna, Tobias Kreiman, Charles Xu, and 1 others. 2024. Octo: An open-source generalist robot policy.arXiv preprint arXiv:2405.12213

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  41. [41]

    Homer Rich Walke, Kevin Black, Tony Z Zhao, Quan Vuong, Chongyi Zheng, Philippe Hansen-Estruch, Andre Wang He, Vivek Myers, Moo Jin Kim, Max Du, and 1 others. 2023. Bridgedata v2: A dataset for robot learning at scale. InConference on Robot Learning (CoRL), pages 1723–1736. PMLR

    2023

  42. [42]

    Jianyuan Wang, Minghao Chen, Nikita Karaev, Andrea Vedaldi, Christian Rupprecht, and David Novotny

  43. [43]

    InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Vggt: Visual geometry grounded transformer. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

  44. [44]

    Jianwei Yang, Reuben Tan, Qianhui Wu, Ruijie Zheng, Baolin Peng, Yongyuan Liang, Yu Gu, Mu Cai, Seonghyeon Ye, Joel Jang, and 1 others. 2025. Magma: A foundation model for multimodal ai agents. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 14203– 14214

    2025

  45. [45]

    Bin Yu, Shijie Lian, Xiaopeng Lin, Zhaolong Shen, Yuliang Wei, Haishan Liu, Changti Wu, Hang Yuan, Bailing Wang, Cong Huang, and 1 others. 2026. 3d-mix for vla: A plug-and-play module for integrating vggt- based 3d information into vision-language-action models.arXiv preprint arXiv:2603.24393

    work page arXiv 2026

  46. [46]

    Bin Yu, Shijie Lian, Xiaopeng Lin, Yuliang Wei, Zhaolong Shen, Changti Wu, Yuzhuo Miao, Xinming Wang, Bailing Wang, Cong Huang, and 1 others. 2026. Twinbrainvla: Unleashing the potential of generalist vlms for embodied tasks via asymmetric mixture-of-transformers.arXiv preprint arXiv:2601.14133

    work page arXiv 2026

  47. [48]

    Xuanran Zhai, Qianyou Zhao, Qiaojun Yu, and Ce Hao. 2025. Vfp: Variational flow-matching policy for multi-modal robot manipulation.arXiv preprint arXiv:2508.01622

    work page arXiv 2025

  48. [49]

    Haoyu Zhen, Xiaowen Qiu, Peihao Chen, Jincheng Yang, Xin Yan, Yilun Du, Yining Hong, and Chuang Gan

  49. [50]

    InInternational conference on machine learning (ICML), pages 61229–61245

    3D-VLA: A 3D vision-language-action generative world model. InInternational conference on machine learning (ICML), pages 61229–61245

  50. [51]

    Jinliang Zheng, Jianxiong Li, Zhihao Wang, Dongxiu Liu, Xirui Kang, Yuchun Feng, Yinan Zheng, Jiayin Zou, Yilun Chen, Jia Zeng, and 1 others. 2025. X-VLA: Soft-prompted transformer as scalable cross-embodiment vision-language-action model.arXiv preprint arXiv:2510.10274

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  51. [52]

    Ruijie Zheng, Yongyuan Liang, Shuaiyi Huang, Jianfeng Gao, Hal Daumé III, Andrey Kolobov, Furong Huang, and Jianwei Yang. 2025. TraceVLA: Visual trace prompting enhances spatial-temporal awareness for generalist robotic policies.arXiv preprint arXiv:2412.10345

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  52. [53]

    Linqing Zhong, Yi Liu, Yifei Wei, Ziyu Xiong, Maoqing Yao, Si Liu, and Guanghui Ren. 2026. Acot-vla: Action chain-of-thought for vision-language-action models.arXiv preprint arXiv:2601.11404

    work page arXiv 2026

  53. [54]

    Zheyuan Zhou, Liang Du, Zixun Sun, Xiaoyu Zhou, Ruimin Ye, Qihao Chen, Yinda Chen, and Lemiao Qiu

  54. [55]

    Main-vla: Modeling abstraction of intention and environment for vision-language-action models.arXiv preprint arXiv:2602.02212

    work page arXiv

  55. [56]

    Brianna Zitkovich, Tianhe Yu, Sichun Xu, Peng Xu, Ted Xiao, Fei Xia, Jialin Wu, Paul Wohlhart, Stefan Welker, Ayzaan Wahid, and 1 others. 2023. Rt-2: Vision-language-action models transfer web knowledge to robotic control. InConference on Robot Learning (CoRL), pages 2165–2183. 15 A Additional Analysis on Intent Consistency and Mode Switching A.1 Mode Swi...

    2023